• Journal of Broadcast Engineering
• /
• v.25 no.7
• /
• pp.1081-1094
• /
• 2020

With the recent development of deep learning, most computer vision-related tasks are being solved with deep learning-based network technologies such as CNN and RNN. Computer vision tasks such as object detection or object segmentation use intermediate features extracted from the same backbone such as Resnet or FPN for training and inference for object detection and segmentation. In this paper, an experiment was conducted to find out the compression efficiency and the effect of encoding on task inference performance when the features extracted in the intermediate stage of CNN are encoded. The feature map that combines the features of 256 channels into one image and the original image were encoded in HEVC to compare and analyze the inference performance for object detection and segmentation. Since the intermediate feature map encodes the five levels of feature maps (P2 to P6), the image size and resolution are increased compared to the original image. However, when the degree of compression is weakened, the use of feature maps yields similar or better inference results to the inference performance of the original image.

• Journal of Korea Multimedia Society
• /
• v.4 no.5
• /
• pp.423-429
• /
• 2001

In this paper, we propose the improved method for retrieving images with DC element of DCT that is used in image compression such as JPEG/MPEG. The existing method retrieves images with DC of fixed block size DCT. In this method, the increase in the block size results in faster retrieving speed, but it lessens the accuracy. The decrease in the block size improves the accuracy, however, it degrades the retrieving speed. In order to solve this problem, the proposed method utilizes the variable block size DCT. This method first determines the existence of object regions within each block, and then creates an image region table. Based on this table, it determines the size of each block, following a simple rule; decrease the block size in the object regions, and increase the block size in the background regions. The proposed method using variable block size DCT improves about 15% in terms of the accuracy. Additionally, when there rarely exist images of same pattern, it is able to retrieve faster only by comparing the image region patterns.

• Journal of Korea Multimedia Society
• /
• v.7 no.2
• /
• pp.173-186
• /
• 2004

This paper proposed the content-based retrieval system as a method for performing image retrieval through the effective feature extraction of the object of significant meaning based on the characteristics of man's visual system. To allow the object region of interest to be primarily detected, the region, being comparatively large size, greatly different from the background color and located in the middle of the image, was judged as the major object with a meaning. To get the original features of the image, the cumulative sum of tile declination difference vector the segment of the object contour had and the signature of the bipartite object were extracted and used in the form of being applied to the rotation of the object and the change of the size after partition of the total length of the object contour of the image into the normalized segment. Starting with this form feature, it was possible to make a retrieval robust to any change in translation, rotation and scaling by combining information on the texture sample, color and eccentricity and measuring the degree of similarity. It responded less sensitively to the phenomenon of distortion of the object feature due to the partial change or damage of the region. Also, the method of imposing a different weight of similarity on the image feature based on the relationship of complexity between measured objects using the fractal dimension by the Boxing-Counting Dimension minimized the wrong retrieval and showed more efficient retrieval rate.

• International conference on construction engineering and project management
• /
• 2022.06a
• /
• pp.1243-1244
• /
• 2022

In recent years, the growing interest in off-site construction has led to factories scaling up their manufacturing and production processes in the construction sector. Consequently, continuous large-scale site monitoring in low-variability environments, such as prefabricated components production plants (precast concrete production), has gained increasing importance. Although many studies on computer vision-based site monitoring have been conducted, challenges for deploying this technology for large-scale field applications still remain. One of the issues is collecting and transmitting vast amounts of video data. Continuous site monitoring systems are based on real-time video data collection and analysis, which requires excessive computational resources and network traffic. In addition, it is difficult to integrate various object information with different sizes and scales into a single scene. Various sizes and types of objects (e.g., workers, heavy equipment, and materials) exist in a plant production environment, and these objects should be detected simultaneously for effective site monitoring. However, with the existing object detection algorithms, it is difficult to simultaneously detect objects with significant differences in size because collecting and training massive amounts of object image data with various scales is necessary. This study thus developed a large-scale site monitoring system using edge computing and a small-object detection system to solve these problems. Edge computing is a distributed information technology architecture wherein the image or video data is processed near the originating source, not on a centralized server or cloud. By inferring information from the AI computing module equipped with CCTVs and communicating only the processed information with the server, it is possible to reduce excessive network traffic. Small-object detection is an innovative method to detect different-sized objects by cropping the raw image and setting the appropriate number of rows and columns for image splitting based on the target object size. This enables the detection of small objects from cropped and magnified images. The detected small objects can then be expressed in the original image. In the inference process, this study used the YOLO-v5 algorithm, known for its fast processing speed and widely used for real-time object detection. This method could effectively detect large and even small objects that were difficult to detect with the existing object detection algorithms. When the large-scale site monitoring system was tested, it performed well in detecting small objects, such as workers in a large-scale view of construction sites, which were inaccurately detected by the existing algorithms. Our next goal is to incorporate various safety monitoring and risk analysis algorithms into this system, such as collision risk estimation, based on the time-to-collision concept, enabling the optimization of safety routes by accumulating workers' paths and inferring the risky areas based on workers' trajectory patterns. Through such developments, this continuous large-scale site monitoring system can guide a construction plant's safety management system more effectively.

• KSCE Journal of Civil and Environmental Engineering Research
• /
• v.29 no.1D
• /
• pp.129-134
• /
• 2009

This study is on the development of photogrammetric methode for measuring the size of object without control points. The model is composed of interior orientation parameters, which are consist of specifications of CCD camera and lens distortion parameters, and exterior orientation parameters, which are calculated through relative orientation and scale adjustment. We evaluated the accuracy of the model to find that it is possible to measure the size of object using the model.

• Journal of Multimedia Information System
• /
• v.6 no.4
• /
• pp.203-208
• /
• 2019

Fast operation of a CNN based object detection is important in many application areas. It is an efficient approach to reduce the size of an input image. However, it is difficult to find an area that includes a target object with minimal computation. This paper proposes a ROI detection method that is fast and robust to noise. The proposed method is not affected by a flicker line noise that is a kind of aliasing between camera and LED light. Fast operation is achieved by using down-sampling efficiently. The accuracy of the proposed ROI detection method is 92.5% and the operation time for a frame with a resolution of 640 × 360 is 0.388msec.

• Journal of the Optical Society of Korea
• /
• v.18 no.6
• /
• pp.698-705
• /
• 2014

We propose a new development for calculating a computer-generated hologram (CGH) through the use of multiple general-purpose graphics processing units (GPGPUs). For optimization of the implementation, CGH parallelization, object point tiling, memory selection for object point, hologram tiling, CGMA (compute to global memory access) ratio by block size, and memory mapping were considered. The proposed CGH was equipped with a digital holographic video system consisting of a camera system for capturing images (object points) and CPU/GPGPU software (S/W) for various image processing activities. The proposed system can generate about 37 full HD holograms per second using about 6K object points.

• Journal of Biomedical Engineering Research
• /
• v.26 no.4
• /
• pp.185-191
• /
• 2005

In this paper, we proposed a downscaled 3D object technique using medical images for telediagnostic use. The proposed system consisted of downscaling/thresholding processes for building a downscaled 3D object and a process for obtaining 2D images at specific angles for diagnosis support. We used 80 slices of Digital Imaging and Communication in Medicine(DICOM) CT images as sample images and the platform-independent Java language for the experiment. We confirmed that the total image set size and transmission time of the original DICOM image set using a down-scaled 3D object decreased approximately $99\%\;and\;98.41\%,$ respectively. With additional studies, the proposed technique obtained from these results will become useful in supporting diagnosis for home and hospital care.

• Journal of the Korea Society for Simulation
• /
• v.12 no.2
• /
• pp.25-34
• /
• 2003

Because of explosive growth in demand for web-based multimedia applications, proper proxy caching for large multimedia object (especially video) has become needed. For a video object which is much larger in size and has different access characteristics than the traditional web object such as image and text, caching the whole video file as a single web object is not efficient for the proxy cache. In this paper, we propose a proxy caching strategy with the constant-sized segment for video file and an improved proxy cache replacement policy. Through the event-driven simulation under various conditions, we show that our proposal is more efficient than the variable-sized segment strategy which has been proven to have higher hit ratio than other traditional proxy cache strategies.

• Proceedings of the KSME Conference
• /
• 2001.06b
• /
• pp.307-312
• /
• 2001

In this paper, we proposed the object-tracking algorithm that recognizes and estimates the any shaped and size objects using vision system. For the extraction of the object from the background of the acquired images, spatio-temporal filter and signature parsing algorithm are used. Specially, for the solution of correspondence problem of the multiple objects tracking, we compute snake energy and position information of the target objects. Through the real-time tracking experiment, we verified the effectiveness of the suggested tracking algorithm.